Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Brigham Young UniversityBYU ScholarsArchive
All Theses and Dissertations
2017-06-01
Perceptual Analysis of Children's Adaptation to anElectropalatography SensorKasey Marie DuffieldBrigham Young University
Follow this and additional works at: https://scholarsarchive.byu.edu/etd
Part of the Communication Sciences and Disorders Commons
This Thesis is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion in All Theses and Dissertations by anauthorized administrator of BYU ScholarsArchive. For more information, please contact [email protected], [email protected].
BYU ScholarsArchive CitationDuffield, Kasey Marie, "Perceptual Analysis of Children's Adaptation to an Electropalatography Sensor" (2017). All Theses andDissertations. 6817.https://scholarsarchive.byu.edu/etd/6817
Perceptual Analysis of Children’s Adaptation to
an Electropalatography Sensor
Kasey Marie Duffield
A thesis submitted to the faculty of Brigham Young University
in partial fulfillment of the requirements for the degree of
Master of Science
Shawn L. Nissen, Chair Christopher Dromey
Kristine Tanner
Department of Communication Disorders
Brigham Young University
Copyright © 2017 Kasey Marie Duffield
All Rights Reserved
ABSTRACT
Perceptual Analysis of Children’s Adaptation to an Electropalatography Sensor
Kasey Marie Duffield
Department of Communication Disorders, BYU Master of Science
The purpose of this study is to observe children’s adaptation to an electropalatographic (EPG) sensor. Sound recordings of six children between the ages of 7;0 and 9;11 sampled at 30-minute intervals over a two-hour period of wearing an EPG sensor were perceptually evaluated to quantify the children’s adaptation over time. Twenty native speakers of American English evaluated the pronunciation of a series of words with embedded stops and fricatives produced with and without an EPG sensor in place. When collapsed over speaker and stimulus type, listener ratings decreased significantly after inserting the EPG sensor. Ratings then increased significantly after the sensor was in place for 30 minutes, and again after 60 minutes. No significant improvement in pronunciation was noted between the 60- and 120-minute test intervals, and adaptation did not reach preplacement levels until the sensor was removed. Mixed results were found in how speakers adapted across the different stimulus types. Adaptation was most consistent across speakers for the conversation conditions, but occurred most rapidly for /s/ and /k/. Speakers showed the best overall adaptation for the phoneme /t/ by the end of testing. These results are similar to several adaptation studies with adults, and the two studies with children. Results from this study will help speech pathologists effectively use EPG technology to help children accurately pronounce speech sounds, and to generalize these pronunciations to their normal speech.
Keywords: electropalatography, EPG, children, adaptation, perceptual analysis
ACKNOWLEDGMENTS
I am grateful for the many people who made this project possible. First, I would like to
thank Dr. Nissen for his efforts advising me through all stages of planning, data collection and
writing for this thesis, and for his creating the computer program used to collect data in this
study. I am also grateful for his allowing me to use his lab, equipment, and software for this
research. Secondly, I would like to acknowledge my committee for their insights, and
suggestions throughout the research process. Thank you to the students who volunteered to
participate as judges for this study, particularly the research assistants who helped pilot the study.
I am grateful to my parents who have encouraged me to continue my education and supported
me as I worked on my thesis. Finally, I am grateful to my husband for his support,
encouragement and insight throughout the thesis process.
iv
TABLE OF CONTENTS
ABSTRACT .................................................................................................................................... ii
ACKNOWLEDGMENTS ............................................................................................................. iii
TABLE OF CONTENTS ............................................................................................................... iv
LIST OF TABLES ......................................................................................................................... vi
LIST OF FIGURES ...................................................................................................................... vii
DESCRIPTION OF THESIS STRUCTURE............................................................................... viii
Introduction ..................................................................................................................................... 1
Method ............................................................................................................................................ 9
Participants .......................................................................................................................... 9
Stimulus .............................................................................................................................. 9
Procedures ......................................................................................................................... 10
Testing for stimulus words.................................................................................... 11
Testing for conversations ...................................................................................... 11
Intra-rater reliability. ............................................................................................. 11
Results ........................................................................................................................................... 12
Time .................................................................................................................................. 12
Time-by-Speaker............................................................................................................... 13
Time-by-Speaker-by-Stimulus Type ................................................................................ 14
Conversation ......................................................................................................... 15
Alveolar fricatives (/s/). ........................................................................................ 22
Palatal fricatives (/ʃ/)............................................................................................. 22
v
Velar stops (/k/)..................................................................................................... 23
Alveolar stops (/t/). ............................................................................................... 23
Discussion ..................................................................................................................................... 24
References ..................................................................................................................................... 29
APPENDIX A: Annotated Bibliography ...................................................................................... 33
APPENDIX B: Consent Form ...................................................................................................... 47
vi
LIST OF TABLES
Table 1 The Mean Rating Each Speaker Received for Each Stimuli at
Each Time Period ..................................................................................................16
vii
LIST OF FIGURES
Figure 1. Average listener rating over time with EPG sensor in place. ............................... 13
Figure 2. Time-by-speaker interaction. ................................................................................ 15
Figure 3a. Average rating of articulation clarity for each stimulus type over time
for Speaker 1 ......................................................................................................... 18
Figure 3b. Average rating of articulation clarity for each stimulus type over time
for Speaker 2. ........................................................................................................ 18
Figure 3c. Average rating of articulation clarity for each stimulus type over time
for Speaker 3. ........................................................................................................ 19
Figure 3d. Average rating of articulation clarity for each stimulus type over time
for Speaker 4. ........................................................................................................ 19
Figure 3e. Average rating of articulation clarity for each stimulus type over time
for Speaker 5. ........................................................................................................ 20
Figure 3f. Average rating of articulation clarity for each stimulus type over time
for Speaker 6. ........................................................................................................ 20
Figure 3g. Average rating of articulation clarity for each stimulus type over time
for Speaker 7. ........................................................................................................ 21
Figure 3h. Average rating of articulation clarity for each stimulus type over time
for Speaker 8. ........................................................................................................ 21
viii
DESCRIPTION OF THESIS STRUCTURE
This thesis was part of a larger collaborative project, portions of which may be submitted
for publication, with the thesis author being one of multiple contributing coauthors. The body of
this thesis was written as a manuscript suitable for submission to a peer-reviewed journal in
speech-language pathology. The analyses conducted in this study were based on a set of
recordings originally collected by Nissen, Celaya, & Knapp (2014). An annotated bibliography is
presented following the reference section in Appendix A. The consent form used in this study is
found in Appendix B.
1
Introduction
Electropalatography (EPG) is an important technology used to track the shape and
movement of the tongue’s contact with the hard palate during speech (Gibbon & Paterson, 2006).
EPG was originally designed by Samuel Fletcher and consisted of an artificial pseudopalate
containing 48 electrodes embedded in a plastic mouthpiece. Information from the electrodes
traveled through a series of external wires and was then projected on a light emitting diode
(LED) display (Fletcher, McCutcheon, & Wolf, 1975). Current EPG sensors, such as the
SmartPalate® produced by CompleteSpeech International® (2015), now have as many as 124
gold-plated electrodes attached to a relatively thin acrylic mouthpiece that is custom molded to
the shape of each user’s upper teeth and hard palate. Information from the sensor is sent through
wires at the front of the mouth to a computer display (CompleteSpeech, 2015). The visual
biofeedback provided by the EPG sensor has been found to be a valuable tool in therapy as it
helps clients to see where their tongue makes contact with the palate, providing visual feedback
on how to adjust their tongue movements to more precisely articulate speech (Carter & Edwards,
2004; Dagenais, 1995; Gibbon & Paterson, 2006; McAuliffe & Cornwell, 2008).
Electropalatography technology is also useful in research, as it helps researchers capture lingua-
palatal contact patterns in real time (Fletcher, et al., 1975).
Electropalatography has been used to describe and treat speech problems associated with
a variety of conditions including cleft palate (Gibbon & Paterson, 2006; Scobbie, Wood, &
Wrench, 2004), articulation impairment (Carter & Edwards, 2004; Dagenais, 1995), and
phonological impairment (Dagenais, 1995; McLeod & Searl, 2006). Motor speech disorders
including apraxia (McAuliffe & Ward, 2006), and dysarthria following traumatic brain injury
(Goozée, Murdoch, & Theodoros, 2003; Kuruvilla, Murdoch, & Goozée, 2008; McAuliffe &
2
Ward, 2006) and dysarthria associated with Parkinson’s disease (McAuliffe, Ward, & Murdoch,
2006) have also been evaluated using EPG.
Although EPG is typically used to treat speech, it can also be used to evaluate and treat
non-speech motor disorders. For example, a study by Mantie-Kozlowski and Pitt (2014) showed
potential for EPG as a tool in helping people with non-speech orofacial myofunctional disorders
(NSOMD) to develop better tongue-to-palate contact and lingual control in swallowing.
Additionally, speech-language pathologists are not the only professionals interested in EPG.
EPG has also been used in research by dentists, orthodontists, and linguists to track the
movement of the tongue (McLeod & Searl, 2006).
A client’s ability to adapt to the EPG device is an important consideration in EPG
treatment. Any structural change to the oral cavity, including foreign objects in the mouth, has
the potential to alter normal speech production. For example, a five-year-old child who loses his
two front teeth or a teenager who was just fitted with an orthodontic retainer may require an
adjustment period as they adapt to the physical changes or obstructions in their mouth
(McFarland, Baum, & Chabot, 1996). Likewise, speakers may need time to adjust to the EPG.
Although relatively thin (1-2 mm) (Cheng, Murdoch, Goozée, & Scott, 2007; Hamlet & Stone,
1982; McAuliffe, Robb, & Murdoch, 2007; McLeod & Searl, 2006), the EPG sensor decreases
the height of the hard palate, thus changing the structure of the oral cavity (McFarland et al.,
1996). Speech pathologists need to be aware of the differences in speech produced by structural
changes resulting from the EPG. Additionally, they need to be aware of the adaptation time
necessary for the client to produce speech that resembles their natural speech as closely as
possible.
3
Previous research on speakers’ adaptation to non-EPG devices such as bite blocks and
lingual magnets may provide insights into how people respond and adapt to a perturbation in the
oral cavity. Bite blocks are material placed between the teeth to set the jaw opening to a fixed
position during speech (McFarland et al., 1996). Speakers usually open and close their mouths
to varying degrees as they speak; however, bite blocks hold the jaw in one place, thus limiting its
range of motion (McFarland et al., 1996). The effects of a bite block on speech may vary
according to phoneme. A study by McFarland and Baum (1995) found that a bite block made
significant changes to the spectral characteristics of vowels, stops, and fricatives immediately
following placement of the bite block. After subjects participated in a 15-minute conversation
with the bite block in place, the formant frequency and duration of vowels returned to normal, or
near-normal levels, but spectral characteristics of consonant productions were still significantly
different from the normal-speaking condition. McFarland and Baum concluded that speakers can
learn to adapt vowel production to a bite block, but consonants require a much longer adaption
time if the speaker is to adapt at all (McFarland & Baum, 1995).
The adaptation of vowel production observed by McFarland and Baum may be explained
by a compensatory behavior called super shaping observed by Gay, Lindblom, and Lubker
(1981) in a study on vowel production in the presence of a bite block. The researchers found the
greatest amount of articulator compensation in areas where the greatest lingua-palatal
constriction occurs in normal speech. That is, the speaker’s efforts were concentrated on these
areas of highest constriction, and, as a result, these areas more closely approximated normal
positioning than did areas of less constriction despite the bite block (Gay et al., 1981).
Lingual magnets are another technology used to track speech movements. Although
lingual magnets are much smaller than bite blocks, they still pose concern for interference with
4
normal speech. However, a study done on the use of a magnetic pellet to track tongue movement
showed no significant difference in spectral characteristics of speech in the fricatives /s/ and /ʃ/
produced with or without the magnet (Dromey, Nissen, Nohr, & Fletcher, 2006). A later thesis
(Hunter, 2016) studied speakers’ adaptation to two sensor coils placed on the midline of the
tongue: one on the tongue tip, and one in the center of the tongue. Three other coils were
attached to the teeth, and upper and lower lips. This study found that perceptual ratings of
articulatory precision decreased after the coils were placed. Acoustic measures showed a shorter
duration following placement of the coils, and the spectral spread and center of gravity increased
for /ʃ/ and decreased for /s/. Differences in acoustic and perceptual measures were fairly constant
after the speakers had worn the sensors for 10 minutes, but did not return to preplacement levels
(Hunter, 2016). The difference in results found in the study by Dromey et al., and Hunter may be
a result of an increased number of sensors used in Hunter’s study.
Although studies on speaker adaptation to bite blocks and lingual magnets provide
valuable insight into how speakers compensate for obstructions in their oral cavity, the structure
and purpose of these devices are different from those of an EPG sensor. Bite blocks are placed
between the teeth to hold the jaw in one position (McFarland et al., 1996), and lingual magnets
are attached to the tongue to track tongue movement in the oral cavity (Dromey et al., 2006).
Electropalatography sensors, on the other hand, cover the hard palate and allow the jaw to move
freely while tracking tongue contact with the palate (McFarland et al., 1996).
Studies on an individual’s ability to adapt to orthodontic retainers are similar to studies
showing the effects of an EPG device. Hamlet and Stone (1982) conducted a study to determine
if a history of lisping was related to difficulty adapting to a dental retainer. The researchers used
an EPG sensor the same size as a dental retainer to record the participants’ articulation patterns
5
before and after wearing the retainer for two weeks. They found that participants did not adapt to
the sensor if they maintained the same articulator adjustments as when the retainer was initially
inserted. The participants who adjusted their articulation throughout the two-week period showed
better adaptation at the end of the study. These results provide some evidence that people adapt
to oral obstructions in different ways, and that some of these adaptations may be more effective
than others (Hamlet & Stone, 1982). Since dental retainers have a similar shape and placement to
EPG sensors, this speaker-by-speaker variation should be considered when evaluating an
individual’s ability to adapt to EPG.
Like dental retainers, palatal appliances resemble the effects of EPG sensors on speech
more closely than technologies such as bite blocks and lingual magnets. Searl, Evitts, and Davis
(2006) described the difference between palatal appliances and EPG devices: “. . . palatal
appliances usually only cover one portion of the hard palate (typically the alveolar ridge), the
thickness is not uniform, and the maximum thickness is often much greater (3-6mm) than that
used in EPG and contact pressure studies” (p. 108). Following the methodology of McFarland
and Baum (1995), McFarland et al. (1996) conducted a study of adaptation to a palatal appliance.
Their findings indicated that the palatal appliance showed little acoustic or perceptual effect on
the production of vowels. Stops showed some acoustic, but not perceptual differences between
speech with the EPG sensor, and speech without. However, the researchers did find perceptual
differences for the fricative /s/ when the palatal appliance was worn. In particular, listeners rated
their ability to identify the sound, and the quality of the sound lower for /s/ immediately after the
palatal appliance was placed; they rated quality, but not their ability to identify the sound, lower
after speakers had participated in a 15-minute conversation. In summary, vowels were the least
affected by a palatal appliance, followed by stops, and then fricatives (McFarland et al., 1996).
6
Several studies have investigated how adults adapt their speech to the presence of an EPG
sensor (McAuliffe et al., 2007; McLeod & Searl, 2006; Searl et al., 2006). Adaptation times
found in these studies varied from 30 minutes to three hours depending on the speech sounds
being evaluated and the criteria for adaptation. McAuliffe et al. used a practise palate—an EPG
without electrodes—to study adults’ adaptation to an EPG sensor. McAuliffe et al., found that
initial imprecision can be detected perceptually immediately after insertion of the device;
however, adaptation typically occurs 45 minutes to three hours postplacement. Additionally,
acoustic measures showed that the segment duration of consonants, and the frequency of vowel
formants are not significantly affected by the practice palate, but the first spectral moment was
reduced for all productions of /s/ regardless of how long the participant had been wearing the
practice palate. Although the practice palate used in this study did not have sensors like an actual
EPG device, the thickness and shape of the practice palate was identical to that of the EPG
sensor. Therefore, studies involving practice palates can be used to predict adaptation to EPG. In
fact, both EPG and practice palates have shown similar changes to the acoustic characteristics of
consonants (McLeod & Searl, 2006).
Searl et al. (2006) studied the acoustic and perceptual effect of an EPG sensor on speech.
In this study, speakers adapted to their EPG sensor in about 30 minutes after placement as
manifest by the acoustic measures returning to preplacement levels. The highest levels of
difference in spectral mean, and stop-gap duration were observed at 15 and 30 minutes
postplacement. Although differences were observed in acoustic measures, they were not manifest
in the perceptual measures. Listeners identified the target phoneme with at least 98 percent
accuracy for every test, and mean distortion ratings were between one and six percent. Searl et
al. claimed that little or no adaptation time is needed for people to produce perceptually
7
undistorted consonants with a 0.5 mm pseudopalate; and acoustically accurate productions occur
after the participant has worn the pseudopalate for about 30 minutes.
In a study on acoustic and perceptual adaptation to an EPG sensor, McLeod and Searl
(2006) found that general adaptation occurred for /t/ after about one hour, and /s/ after two hours;
however, no speaker produced speech that acoustically matched the no-palate condition. McLeod
and Searl noted that overall, the palate had only a small effect on perceptual distortion ratings,
with affricates and fricatives being most notably affected. Findings from McLeod and Searl
differ from the study by Searl et al. (2006) which reported that acoustic measures reached
preplacement levels. According to these studies on adult adaptation to EPG, adaptation can occur
anywhere between 30 minutes and three hours after the palate is inserted. The adaptation time
can vary depending on phoneme and speaker. Additionally, adaptation time can vary based on
adaptation criteria (i.e., acoustic or perceptual); that is, a subject’s rate of adaptation as judged by
a listener is typically recorded as happening more quickly and completely than the acoustic
measures of adaptation for the same subject (McAuliffe et al., 2007; McLeod & Searl, 2006;
Searl et al., 2006).
Most EPG adaptation studies have described how adults adapt to the presence of the EPG
sensor; however, children are also an important population that can benefit from EPG therapy. A
2006 study by Gibbon and Paterson surveyed 10 speech-language therapists about the use of
EPG in therapy. Therapists were asked to include information concerning demographics, type of
disorder, therapy given, and effect of EPG on the client’s progress. Of the 60 clients that the
therapists reported as using EPG in therapy, 51 percent began therapy with an EPG device
between the ages of six and ten and 81 percent started before the age of 15 (Gibbon & Paterson,
2006).
8
Although children make up a large proportion of EPG users, relatively little research has
examined how younger speakers adapt their speech to the presence of the sensor. Two
unpublished theses by Knapp (2014) and Celaya (2014) examined children’s adaptation to an
EPG sensor by evaluating the acoustic characteristics of stop and fricative productions. Six
children were recorded saying words containing the sounds /t/, /k/, /s/, and /ʃ/ without the EPG
sensor in place, at five, 30-minute intervals while wearing the EPG sensor, immediately after the
sensor was removed, and 30 minutes following removal. Knapp (2014) noted that for the
phonemes /t/ and /k/, acoustic measures at the end of testing with the sensor (two hours
postplacement) resembled preplacement levels for three of the six participants; two of the six
participants showed partial adaptation to the sensor. An acoustic analysis of /s/ and /ʃ/ completed
by Celaya indicated that the child speakers adapted to the EPG sensor within 30 minutes to two
hours. The timing and consistency of adaptation varied by spectral characteristic. Fricative
duration was least affected by the presence of the EPG sensor, whereas the fricative intensity was
often impacted by the sensor, a decrease after placement of the EPG followed by a gradual
increase throughout the two hours. Spectral mean and variance were both affected upon
placement of the EPG sensor, and some participants began to adapt by 30 minutes
postplacement; however, measures of mean and variance were more variable across speakers,
and adaptation was less consistent across the two hours of testing than for spectral intensity
(Celaya, 2014).
The work by Knapp (2014) and Celaya (2014) provides insight into how children might
adapt to an EPG sensor in terms of speech acoustics; however, little research has examined
children’s adaptation in terms of the perceptual salience of their speech production. Thus, the
9
current thesis seeks to expand the previous research of Knapp (2014) and Celaya (2014) by using
a perceptual analysis to quantify children’s adaptation to EPG over time.
Method
Participants
Participants for this study consisted of 25 university students enrolled in the Department
of Communication Disorders at Brigham Young University (BYU). Data from four participants
was excluded from the study due to an intra-rater reliability coefficient of less than .50, and one
participant for limited linguistic experience. Before testing, participants’ hearing was screened
via pure-tone air-conduction testing at 25 dB across one-octave intervals from 500-8000 Hz. In
addition, all participants signed a consent form prior to testing. Procedures for this study have
been reviewed and approved by the institutional review board at BYU.
Stimulus
Stimulus items for this study were extracted from audio recordings collected by Knapp
(2014) and Celaya (2014). The target stimuli in these studies were produced by eight children
between the ages of 7;0 and 9;11 years. Speakers were asked to say target words in the carrier
phrase Say ___ again three times, and then to respond to a question asked by the examiner.
Speakers completed this task at eight intervals throughout a three-hour session: (a) prior to
placement of an EPG sensor, (b) directly after inserting the EPG sensor, (c) after 30 minutes of
wearing the sensor, (d) after 60 minutes, (e) after 90 minutes, (f) after 120 minutes, (g)
immediately after the sensor was removed, and (h) 30 minutes after removing the sensor. All
audio recordings were collected in a sound-attenuating booth using a high-quality, low-
impedance dynamic microphone and preamplifier. Recordings were made with a sampling rate
of 44.1 kHz and a quantization of 16 bits.
10
Stimulus items in the current study included a portion of the recordings collected by
Knapp and Celaya. The speech samples consisted of 768 individual sentence files (eight
participants x eight test intervals x four stimulus words x repeated three times). The stimulus
words included the sounds /t/, /k/, /s/, and /ʃ/ in the initial position followed by a high-front
vowel. One of the three repetitions of each word was randomly selected for perceptual analysis
resulting in a total of 256 words. Stimuli also included 64 conversation samples (eight
participants x eight test intervals).
In preparation for perceptual analysis, words were extracted from the carrier phrases
using Adobe Audition (Version 9). Then words were normalized for intensity, and filtered to
extract any electronic noise and noise artifacts. Additionally, 500 ms of silence was added before
and after each word. The conversation samples were modified to remove the examiner’s
comments and edited to be approximately 30 seconds long, with an electronic beep signaling the
end of each sample.
Procedures
The listeners evaluated the extracted speech samples in one 60-minute session, divided
into three test periods. They evaluated the stimuli with word-initial fricatives in a 15-minute test
period, stimuli with word-initial stops in another 15-minute period, and the conversation samples
in a 30-minute period. The three tests were administered in a random order, and participants
were offered a two-minute break after 30 minutes. The session began with a hearing screening
and instructions.
Audio signals for both tests were presented to the participants via headphones. The
participants were allowed to select a comfortable intensity for the stimuli with a starting level of
approximately 60 dB HL. The system did not permit intensities outside the range of safe hearing.
11
Before presenting the test stimuli, each participant evaluated a practice trial of each stimulus type
to ensure that they understood the rating system and that the equipment was adjusted properly.
Testing for stimulus words. Participants were instructed to listen to the initial consonant
of each word presented and then rate their perception of the correctness of the consonant
production using a sliding analog scale from 0 to 100 (0 corresponding to a completely distorted
production and 100 corresponding to a typical, undistorted production). Participants were
provided with written instructions including the two words they would hear for that particular
test, and directions for using the scale. Using a custom computer program, all recordings were
presented to participants in a random order. After rating a production, participants were
instructed to advance to the next stimulus item by using the mouse to select a small box on a
computer screen. Participants were informed that they could replay an item if they missed
hearing the stimuli due to an external distraction or technical error, but the test was not designed
for multiple repetitions of each stimulus.
Testing for conversations. Participants were instructed to listen to each conversation,
and then rate the overall intelligibility of the conversation from 0 to 100 (0 corresponding to
completely unintelligible, and 100 corresponding to typical, intelligible speech) by using the
same analog sliding scale as in the previous session. Participants were instructed to wait for a
beep signaling the end of the conversation before proceeding to the next item. After rating the
conversation, participants were instructed to advance to the next stimulus item by using the
mouse to select a small box on the computer screen.
Intra-rater reliability. Ten percent of the stimulus items were retested for each listener.
The ratings for the first and second presentation of the stimulus were compared using a Pearson
Correlation. Four participants received a correlation of less than 0.5 and were excluded from the
12
study. There was a correlation [r2 = .817, p = 0.01] between the scores of the remaining
participants indicating that the participants included in this study showed statistically significant
reliability for the ratings they provided.
Results
Inferential statistics for this study involved a repeated-measures analysis of variance
(ANOVA) with three within-subject factors (speaker, time period, and stimulus type). The
ANOVA results include a measure of effect size, partial eta squared, or η2. The value of this
power statistic (η2) can range from 0.0 to 1.0, and is considered a proportion of variance
explained by a dependent variable when controlling for other factors. Greenhouse-Geisser
adjustments were used to adjust the F-tests with regard to the degrees of freedom when
significant deviations from sphericity were present. In addition, pairwise comparisons for
significant within-subject factors were calculated using General Linear Model repeated-measures
contrasts with associated F-tests.
Time
According to the results of the ANOVA, when the listener ratings were collapsed across
speaker and stimulus type, the difference in average ratings between time periods was
statistically significant, F(7, 133) = 169.96, p < .001, η2 = .90. As illustrated in Figure 1, there
was an overall decrease in pronunciation clarity upon insertion of the EPG sensor, with a gradual
increase in clarity until 60 minutes postplacement. At 60 minutes, the improvement plateaued
below preplacement levels. Ratings returned to preplacement levels once the EPG device was
removed. Results of the pair-wise comparison support the trend illustrated in the graph. The
comparison showed a significant difference between the ratings before and immediately after the
EPG was inserted, p < .001, with a mean difference of 34.52. Listener ratings increased
13
significantly from the time the EPG was placed, and 30 minutes postplacement, p < .001, with a
mean difference of -7.17. Ratings continued to increase between 30 and 60 minutes
postplacement, p < .001, with a mean difference of -6.39. The difference between tests at 60, 90
and 120 minutes postplacement was not significant, suggesting that adaptation to the EPG
plateaued, or stopped after 60 minutes. Ratings did not increase again until the EPG was
removed, as indicated by a significant difference between ratings 120 minutes postplacement and
immediately after removal of the sensor, p < .001. The pair-wise comparison also showed no
significant difference among the three tests with the sensor removed.
Figure 1. Average listener rating over time with EPG sensor in place. The average listener rating of articulatory precision on a scale from 0 to 100 over the eight time periods. Ratings are collapsed across speaker and stimulus type.
Time-by-Speaker
The ANOVA also indicated a significant interaction between the listener ratings at each
time period and the individual speakers, F(49,931) = 30.28, p < .001, η2 = .61. As shown in
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Rat
ing
of A
rtic
ulat
ory
Prec
ision
(0-
100)
Test Period (minutes)
14
Figure 2, all speakers followed the pattern of an initial decrease in speech clarity following EPG
placement; however, the magnitude of this decrease varied depending on the speaker. For
example, Speaker 5 dropped 71 average rating points upon insertion of the EPG sensor
(Preplacement = 95.2; Time Period 0 = 24.0), while Speaker 6 dropped only 11 points between
the preplacement condition (86.3) and immediately after the palate was placed (75.3). Five of the
eight speakers showed some adaptation, as manifest by an increase in listener ratings by 30
minutes postplacement, while ratings for the other three speakers continued to decrease until 60
minutes postplacement.
Patterns for listener ratings between 60 and 120 minutes postplacement varied by
individual speaker. Ratings for Speaker 2 plateaued, and ratings for Speaker 3 gradually
increased between 60 and 120 minutes of wearing the palate—similar to the average pattern seen
when observing ratings collapsed over speaker and stimulus type. The other six speakers showed
no additional improvement in the clarity of their pronunciation after 90 minutes of wearing the
sensor.
The extent to which each speaker adapted to the EPG sensor varied as well. Ratings for
Speaker 5 increased by 22 points over the time the EPG was in place (Time Period 0 = 24.0 and
Time Period 120 = 46.3). On the other hand, Speaker 6 gained only 2.8 points while the EPG
was in place (Time Period 0 = 75.3 and Time Period 90 = 78.1). Despite these differences, all
speakers returned to preplacement levels after removal of the EPG sensor.
Time-by-Speaker-by-Stimulus Type
Results from the ANOVA showed that the time-by-speaker ratings also varied as a
function of the stimulus type, F(196, 3724) = 6.17, p < .001, η2 = .25. Table 1 contains a detailed
15
listing of the listener ratings across speaker, time period, and stimulus type. An illustration of
these differences is shown in Figures 3a-3h.
Conversation. Average ratings were the most consistent across speakers for the
conversation sample condition. Figures 3a-h illustrate that most speakers showed a drop in
ratings when the sensor was inserted, followed by a gradual increase while the palate was in
place and a return to preplacement clarity when it was removed. The only exception was Speaker
5 (Figure 3e) whose ratings did not increase over the two-hour adaptation period. Speakers 3, 6,
and 8 reached preplacement levels at least once while wearing the EPG sensor.
Figure 2. Time-by-speaker interaction. The average listener ratings collapsed across stimulus types for each speaker at each time period.
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Test Period (minutes)
Speaker 1
Speaker 2
Speaker 3
Speaker 4
Speaker 5
Speaker 6
Speaker 7
Speaker 8
16
Table 1
The Mean Rating Each Speaker Received for Each Stimuli at Each Time Period
Speaker Stim.a Pre 0 30 60 90 120 Post Post +30
1 Conv. 81.0 33.7 38.9 58.0 56.1 44.6 83.0 82.4
/k/ 93.5 27.5 63.8 55.5 64.9 46.0 91.8 93.5
/t/ 94.9 47.6 53.2 47.1 61.6 63.6 75.1 76.3
/s/ 87.2 15.6 72.3 41.5 68.5 57.1 87.8 87.7
/ʃ/ 80.7 49.3 21.0 35.0 50.3 41.5 86.6 81.9
2 Conv. 78.1 30.5 36.9 46.3 61.5 59.1 80.2 72.9
/k/ 65.5 51.9 78.0 64.5 61.0 76.2 91.7 89.0
/t/ 84.3 77.8 55.7 86.8 63.6 57.1 95.5 93.1
/s/ 91.5 86.9 76.5 94.1 94.3 91.2 85.9 86.2
/ʃ/ 90.1 82.6 55.6 90.3 87.5 87.0 93.4 94.4
3 Conv. 89.4 49.4 41.7 47.8 59.0 87.3 89.7 89.8
/k/ 91.6 70.1 63.6 70.7 73.0 74.9 92.7 94.2
/t/ 89.1 41.7 69.8 80.2 82.2 86.3 96.3 94.8
/s/ 89.3 56.4 56.5 82.6 83.1 75.1 94.0 92.0
/ʃ/ 69.6 47.3 49.2 72.0 80.9 59.0 91.6 91.1
4 Conv. 85.9 31.1 52.8 65.9 58.4 64.6 89.6 84.1
/k/ 95.3 38.2 91.7 90.7 75.3 86.2 89.8 94.6
/t/ 94.0 67.1 93.2 80.1 71.9 95.0 92.9 95.7
/s/ 89.8 33.3 88.4 74.1 68.0 75.8 79.2 88.3
/ʃ/ 91.3 32.8 85.8 79.2 69.7 74.1 91.5 91.6
5 Conv. 93.6 34.1 30.4 40.7 27.0 30.3 95.7 87.7
/k/ 93.4 23.8 25.3 36.9 39.3 46.3 88.7 90.2
/t/ 98.6 13.1 38.1 62.6 49.8 64.4 93.4 96.8
/s/ 97.0 18.4 39.1 32.1 19.6 21.3 83.8 93.9
/ʃ/ 93.2 30.5 48.9 59.2 43.4 42.9 91.5 88.9
17
Cont.
Speaker Stim.a Pre 0 30 60 90 120 Post Post +30
6 Conv. 81.0 63.6 76.1 70.2 80.3 77.5 83.9 85.7
/k/ 93.8 78.3 83.4 49.5 85.4 78.4 85.6 92.0
/t/ 82.5 87.7 95.7 85.8 92.6 60.7 86.0 91.0
/s/ 88.2 79.9 50.7 80.5 63.5 69.0 90.5 79.9
/ʃ/ 86.3 66.9 48.3 51.1 68.7 70.2 93.0 82.7
7 Conv. 91.1 57.7 52.6 71.0 79.0 76.2 88.5 83.6
/t/ 88.3 63.6 34.3 65.7 78.3 69.6 91.7 91.7
/k/ 89.3 66.3 60.4 67.2 69.5 56.5 71.5 70.6
/s/ 77.7 37.9 42.4 78.1 43.3 53.9 82.2 72.6
/ʃ/ 74.8 62.8 59.6 77.1 72.1 57.7 85.6 87.9
8 Conv. 84.7 58.0 58.2 59.1 70.9 79.2 82.2 92.9
/k/ 83.6 66.5 77.7 89.9 77.4 86.6 92.3 89.0
/t/ 57.0 69.7 61.4 35.2 68.8 49.2 80.8 52.7
/s/ 84.6 59.5 70.5 77.5 74.9 73.9 81.4 88.7
/ʃ/ 71.4 52.1 50.4 51.6 84.6 68.9 70.1 74.4
Note. aConv. = listener ratings of pronunciation clarity for the 30-second conversation samples.
18
Figure 3a. Average rating of articulation clarity for each stimulus type over time for Speaker 1.
Figure 3b. Average rating of articulation clarity for each stimulus type over time for Speaker 2.
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
19
Figure 3c. Average rating of articulation clarity for each stimulus type over time for Speaker 3.
Figure 3d. Average rating of articulation clarity for each stimulus type over time for Speaker 4.
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
20
Figure 3e. Average rating of articulation clarity for each stimulus type over time for Speaker 5.
Figure 3f. Average rating of articulation clarity for each stimulus type over time for Speaker 6.
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y P
reci
sion
(0-1
00)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
21
Figure 3g. Average rating of articulation clarity for each stimulus type over time for Speaker 7.
Figure 3h. Average rating of articulation clarity for each stimulus type over time for Speaker 8.
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
0
10
20
30
40
50
60
70
80
90
100
Pre 0 30 60 90 120 Post Post +30
Aver
age
Ratin
g of
Art
icul
ator
y Pr
ecisi
on (0
-100
)
Time Period (minutes)
Conv.
/k/
/t/
/s/
/ʃ/
22
Alveolar fricatives (/s/). Speakers showed an increase in listener ratings for /s/ more
quickly than for the other sounds. Six speakers showed an increase in ratings between 0 and 30
minutes of wearing the sensor; however, five of these six showed a decrease in speech clarity
during the adaptation period at 60, 90 or 120 minutes. Although listener ratings for Speaker 2
(Figure 3b) initially decreased from 86.9 to76.5 during the first 30 minutes of wearing the EPG
sensor, his scores plateaued around his preplacement value (91.5) with ratings of 94.1, 94.3, and
91.2 for tests at 60, 90, and 120 minutes, respectively. Ratings for Speaker 8 (Figure 3h) also
plateaued after the palate was in place for 60 minutes, but his ratings were lower than
preplacement levels. Four speakers reached preplacement ratings at least once while the EPG
was in place.
Palatal fricatives (/ʃ/). Individual speaker ratings varied more for /ʃ/ than for /s/ and the
conversation sample. The perceptual ratings for most speakers decreased, or remained low
between 0 and 30 minutes following insertion of the EPG sensor (Figure 3a-h). Improvement
took longer than for /s/ with six speakers making some improvements by 60 minutes. However,
like for /s/, the ratings for these speakers decreased again at 90 and/or 120 minutes. The two
exceptions were Speaker 2 (Figure 3b) whose ratings remained near preplacement levels for the
last hour of wearing the device, and Speaker 6 speaker (Figure 3f) whose ratings gradually
increased from 30 to 120 minutes of wearing the palate. Four speakers reached preplacement
clarity at least once while the EPG was in place.
Except for Speaker 8 (Figure 3h), all speakers showed similar or near-similar trends for
/s/ and /ʃ/. Three speakers (Figures 3b-d) demonstrated the same trend for /s/ and /ʃ/ at all time
periods with /ʃ/ rated less clear than /s/. As seen in Figures 3e and 3f, Speakers 5 and 6 showed
the same trend in ratings for /s/ and /ʃ/ except for 60 minutes postplacement. Speakers 1 and 7
23
(Figures 3a and 3g) showed similar trends for three of the five adaptation periods. All but two
speakers showed generally lower listener ratings for /ʃ/ than for /s/. Listener ratings for speakers
2, 7, and 8 reached preplacement levels at least once for each sound.
Velar stops (/k/). Like /s/, most speakers showed improvement by 30 minutes
postplacement for /k/; however, fewer reached preplacement speech clarity while the EPG was in
place. For stimulus /k/, five speakers reached a plateau by 30 minutes postplacement; however,
all speakers showed one reduction in listener ratings for one test at 60, 90 or 120 minutes.
Speakers 2, 4 and 8 were the only speakers to near preplacement listener ratings with the EPG in
place. As seen in Figures 3c and 3e, listener ratings for Speaker 3 and Speaker 5 began to
increase gradually by 30 minutes postplacement, but neither achieved preplacement speech
clarity with the EPG sensor in place.
Alveolar stops (/t/). As illustrated in Figures 3a-h, listener ratings for /t/ showed more
speaker variability than the other stimulus types. Speakers 1 and 3 had similar trends with a
decrease in speech clarity following placement of the EPG, and a gradual increase in clarity over
time. While Speaker 3 reached near-normal clarity by 120 minutes, Speaker 1 did not. The trends
in the ratings for the remaining speakers were inconsistent and distinct. Similar to /s/ and /ʃ/, four
speakers reached preplacement levels at least once following placement of the EPG sensor, but
none of them maintained preplacement speech clarity. For example, Speaker 4 (Figure 3d) had a
preplacement rating of 94.0, he received a similar rating at 30 minutes (93.2) and 120 minutes
(95.0), but not for the other time periods. On the other hand, Speaker 6, as shown in Figure 3f,
increased in clarity from 82.5 preplacement to 87.7 following placement, and 95.7 after 30
minutes. He maintained this clarity until 120 when his score dropped to 60.7. Speaker 7 (Figure
3g) received a rating of 89.3 prior to insertion of the EPG, but the remainder of his scores—
24
including scores after the EPG was removed—were between 56.5 and 71.5. Despite the
variability among speakers, listener ratings were concentrated above 50 for the test periods 90
and 120 minutes postplacement. Ratings were not concentrated above 50 while the EPG was in
place for any other stimulus type. Six of the speakers showed similar trends for /k/ and /t/, for at
least seven of the eight time periods. That is, the speakers generally followed the same trend for
improvement or regression, but the magnitude of the change was different depending on the
stimulus type.
Discussion
When collapsed over speaker and stimulus type, there was a significant difference in
listener ratings before and after inserting the EPG sensor. Ratings also increased significantly
after talking with sensor in place for 30 minutes, and again after talking with the sensor for an
additional 30 minutes. No significant improvement in pronunciation was noted between the 60-
and 120-minute test intervals. Overall, the speakers’ pronunciation significantly improved after
wearing the sensor for 30 minutes, and after 60 minutes; however, as a group their pronunciation
did not reach preplacement levels until the sensor was removed.
When collapsed across stimulus type, individual speakers differed in how well they
adapted to the EPG sensor; some speakers reached preplacement ratings with the EPG in place,
while others did not. Additionally, adaptation patterns varied with some speakers showing
improvement by 30 minutes and others by 60 minutes, with most speakers showing no additional
improvement after wearing the sensor for 90 minutes.
Speakers showed mixed results in how they adapted across the different stimulus types.
For the conversation sample, most speakers showed no improvement after 90 minutes following
EPG placement, and only one speaker completely adapted to the EPG sensor. The speakers’
25
adaptation patterns were more variable across the different consonants compared to their
performance in conversation. Ratings increased the most rapidly for /s/ and /k/ with improvement
beginning after 30 minutes of wearing the sensor. Although ratings for /t/ did not improve as
rapidly as they did for /s/ and /k/, they were higher than ratings for the other stimuli by the end of
testing with the EPG in place. This indicates that participants approached full adaptation for /t/
more than for the other sounds.
The individual speaker differences observed in the current study are similar to those seen
by Hamlet and Stone (1982). They observed that all participants made adjustments such as
tongue retraction or advancement, groove narrowing, or jaw placement with initial placement of
a dental retainer. Those who adapted to the retainer showed a change in articulator placement
following the two-week period, indicating that they altered their articulator placement during the
adaptation time. Those participants who did not adapt showed no change. In the current study,
speakers differed in their adaptation patterns, regardless of the stimulus type. These individual
differences may be the result of the same articulatory changes made by participants in the study
by Hamlet and Stone.
Although Hamlet and Stone (1982) observed the effect of individual speaker differences
on adaptation, several other adaptation studies of adults addressed differences across stimulus
types. McFarland et al. (1996) found that stops showed no perceptual differences in speech
immediately after placing the EPG sensor, or following a 15-minute conversation. Perceptual
ratings for the fricative /s/ remained low following the conversation. Although listeners in the
current study heard a decrease in quality of stops, the speakers’ adaptation patterns for /t/ were
generally better than for fricatives. The acoustic findings of a study by McAuliffe et al. (2007)
are consistent with differences in perceptual clarity between /t/ and the fricatives /s/ and /ʃ/. In an
26
acoustic analysis, they found that the segment duration of consonants, and the frequency of
vowel formants did not change with placement of the EPG sensor, while the EPG did change the
first spectral moment for /s/.
Searl et al. (2006) used both an acoustic and a perceptual analysis to study the time
needed for adult speakers to adapt their speech to a pseudopalate when producing the phonemes
/t/ and /s/. Acoustic measures changed upon initial placement of the pseudopalate, but returned to
normal by 30 minutes. For the perceptual analysis, listeners reported no differences in
pronunciation of /t/ and /s/ following placement of the pseudopalate. This differs from the
current study where listeners did report differences in speech clarity for both /t/ and /s/ following
placement of the EPG sensor. One reason may be a difference in the scale provided to the
listeners. The current study asked listeners to rate clarity; Searl et al. asked listeners to rate
distortion. Additionally, adult speakers may adapt to the EPG more efficiently than children
because adults have fully-developed speech, and because their mouth is proportionally larger as
compared to the EPG sensor.
Although Searl et al. (2006) did not find that an adaptation period was necessary for
speakers to reach preplacement speech production, a study by McAuliffe et al. (2007) found that
adult speakers did need between 45 minutes and 3 hours to fully adapt their speech to the
presence of the EPG sensor. Of note, ratings for the phoneme /s/ adapted the most quickly with a
50 percent increase in listener ratings between the initial placement of the sensor and 45 minutes
postplacement, followed by a return to near-normal listener ratings by three hours of wearing the
sensor. Similarly, in the current study, listener ratings of the children’s speech pronunciation
clarity increased by 30 minutes postplacement for the phoneme /s/; this is faster than they did for
/t/ and /ʃ/.
27
In two previous graduate theses, Knapp (2014) and Celaya (2014) performed an acoustic
analysis of the speech samples used in this study. Overall, they found that the duration was least
affected by the presence of the EPG sensor. Similar to the current study, measures of intensity,
showed a decrease after placement of the EPG followed by a gradual increase throughout the two
hours, but intensity did not return to normal while the EPG was in place. Likewise, measures of
spectral mean for /t/ and /ʃ/ showed a change with insertion of the EPG followed by general
improvement. On the other hand, measures of spectral mean for /s/ and /k/ were inconsistent
across speakers. Knapp and Celaya also observed that spectral variance was the most variable
acoustic characteristics across speakers (Celaya, 2014) which is consistent with the individual
variability observed in the current study.
Several limitations of this study should be addressed in future research. One was that a
number of children gave short responses to the conversation prompts, while others’ responses
were more comprehensive. Speech rate also varied across speakers and test periods. In future
studies, researchers could help children regulate their rate of speech across test periods by
displaying the cue cards at a consistent rate, and by encouraging the children to use a pacing
board. Additionally, researchers could maximize the length of the conversation by asking the
children to talk for a certain amount of time, and minimize clinician interruption during the
conversation. Another limitation is possible test fatigue which may have led the children to
perform more poorly on the tests at 60, 90 and 120 minutes postplacement. Some of the speakers
may have grown tired because of the length of the study, or because the EPG device was
uncomfortable, or because of the repetitive nature of the test stimuli over the eight test intervals.
To minimize the effects of test fatigue, researchers should consider shortening the overall test
time, or the amount of stimuli. An incentive after each test, such as the opportunity to move on to
28
a new activity or earn a prize, would also motivate the children to increase their focus during
each test.
Despite these limitations, findings of the current study provide insight into how children
adapt to an EPG sensor and how their adaptation differs from that of adults. This information is
particularly useful to clinicians who use EPG with their clients. When using EPG therapy,
clinicians may benefit from knowing that individual differences can influence how quickly and
completely a child adapts to an EPG sensor, and thus tailor treatment to each client. Additionally,
adaptation across speakers and sounds is generally neither immediate, nor complete; as such, a
client’s speech productions will not sound like normal speech as long as the sensor is in place.
As a result, clinicians may want to give children the opportunity to practice speech sounds
following EPG treatment. Doing so will increase the effectiveness of EPG treatment by allowing
children to generalize what they learned during EPG therapy to their everyday speech.
29
References
Adobe Audition CC (Version 9). [Computer Software]. San Jose, CA: Adobe Systems
Incorporated.
Carter, P., & Edwards, S. (2004). EPG therapy for children with long-standing speech disorders:
Predictions and outcomes. Clinical Linguistics and Phonetics, 13, 359-372.
doi:10.1080/02699200410001703637
Celaya, M., (2014). Speech adaptation to electropalatography in children’s productions of /s/ and
/ʃ/ (master’s thesis). Retrieved from BYU Scholars Archive. Paper No. 4103.
Cheng, H. Y., Murdoch, B. E., Goozee, J. V., & Scott, D. (2007). Electropalatographic
assessment of tongue-to-palate contact patterns and variability in children, adolescents,
and adults. Journal of Speech, Language, and Hearing Research, 50, 375-392. doi:1092-
4388/07/5002-0375
Complete Speech International© (2015). Smart Palate [apparatus and software].Orem, Utah,
USA.
Dagenais, P. A. (1995). Electropalatography in the treatment of articulation/phonological
disorders. Journal of Communication Disorders, 28, 303-329. doi:10.1016/0021-
9924(95)00059-1
Dromey, C., Nissen, S., Nohr, P., & Fletcher, S. G. (2006). Measuring tongue movements during
speech: Adaptation of a magnetic jaw-tracking system. Speech Communication, 48, 463-
473. doi:10.1016/j.specom.2005.05.003
Fletcher, G. S., McCutcheon, M. J., & Wolf, M. B. (1975). Research note: Dynamic palatometry.
Journal of Speech and Hearing Research, 18, 812-819. doi:10.1044/jshr.1804.812
30
Gay, T., Lindblom, B., & Lubker, J. (1981). Production of bite-block vowels: Acoustic
equivalence by selective compensation. Journal of the Acoustical Society of America, 69,
802-810. doi: 10.1121/1.385591
Gibbon, F. E., & Paterson, L. (2006). A survey of speech and language therapists’ views on
electropalatography therapy outcomes in Scotland. Child Language Teaching and
Therapy, 22, 275-292.
Goozée, J. V., Murdoch, B. E., & Theodoros, D. G. (2003). Electropalatographic assessment of
tongue-to-palate contacts exhibited in dysarthria following traumatic brain injury: Spatial
characteristics. Journal of Medical Speech-Language Pathology, 11, 115-129.
Hamlet, S. L., & Stone, M. (1982). Speech adaptation to dental prosthesis: The former lisper.
The Journal of Prosthetic Dentistry, 47, 564-478. doi: 10.1016/0022-3913(82)90311-0
Hunter, E. (2016). Speech adaptation to kinematic recording sensors (master’s thesis). Retrieved
from BYU Scholars Archive. Paper No. 5728.
Jongman, A., Wayland, R., & Wong, S. (2000). Acoustic characteristics of English fricatives.
Journal of the Acoustical Society of America, 108, 1252-1263. doi: 10.1121/1.1288413
Knapp, K. B., (2014). Children’s adaptation to electropalatography: Evidence from acoustic
analysis of /t/ and /k/ (master’s thesis). Retrieved from BYU Scholars Archive. Paper No.
4118.
Kuruvilla, M. S., Murdoch, B. E., & Goozée, J. V. (2008). Electropalatographic (EPG)
assessment of tongue-to-palate contacts in dysarthric speakers following TBI. Clinical
Linguistics & Phonetics, 22, 703-725. doi: 10.1080/02699200802176378
31
Mantie-Kozlowski, A., & Pitt, K. (2014). Treating myofunctional disorders: A multiple-baseline
study of a new treatment using electropalatography. American Journal of Speech-
Language Pathology, 23, 520-529. doi:10.1044/2014_AJSLP-14-0001
McAuliffe, M. J., & Cornwell, P. L. (2008). Intervention for lateral /s/ using electropalatography
(EPG) biofeedback and an intensive motor learning approach: A case report.
International Journal of Language and Communication Disorders 43, 219-229. doi:
0.1080/13682820701344078
McAuliffe, M. J., Robb, M. P., & Murdoch, B. E. (2007). Acoustic and perceptual analysis of
speech adaptation to an artificial palate. Clinical Linguistics and Phonetics 21, 885-894.
doi: 10.1080/02699200701576827
McAuliffe, M. J., & Ward, E. C. (2006). The use of electropalatography in the assessment and
treatment of acquired motor speech disorders in adults: Current knowledge and future
directions. NeuroRehabilitation, 21, 189–203. Retrieved from http://www.ncbi .nlm.nih.gov
/pubmed/17167188
McAuliffe, M. J., Ward, E. C., & Murdoch, B. E. (2006). Speech production in Parkinson's
disease: II. Acoustic and electropalatographic investigation of sentence, word and segment
durations. Clinical Linguistics & Phonetics, 20, 19-33. doi: 10.1080/02699200701576827
McFarland, D. H., & Baum, S. R. (1995). Incomplete compensation to articulatory perturbation.
Journal of the Acoustical Society of America, 97, 1865-1873. doi: 10.1121/1.412060
McFarland, D. H., Baum, S. R., & Chabot, C. (1996). Speech compensation to structural
modifications of the oral cavity. Journal of the Acoustical Society of America, 100, 1093-
1104. doi: 10.1121/1.416286
32
McLeod, S., & Searl, J. (2006). Adaptation to an electropalatograph palate: Acoustic,
impressionistic, and perceptual data. American Journal of Speech-Language Pathology,
15, 192-206. doi:10.1044/1058-0360(2006/018)
Scobbie, J. M., Wood, S. E., & Wrench, A. A. (2004). Advances in EPG for treatment and
research: An illustrative case study. Clinical Linguistics & Phonetics, 18, 373-389.
Retrieved from http://search.proquest.com/docview/85406988?accountid=4488
Searl, J., Evitts, P., & Davis, W. J. (2006). Perceptual and acoustic evidence of speaker
adaptation to a thin pseudopalate. Logopedics Phoniatrics Vocology, 31, 107-116. doi:
10.1080/14015430500390961
33
APPENDIX A: Annotated Bibliography
Cheng, H. Y., Murdoch, B. E., Goozee, J. V., & Scott, D. (2007). Electropalatographic assessment of tongue-to-palate contact patterns and variability in children, adolescents, and adults. Journal of Speech, Language, and Hearing Research, 50, 375-392. doi:1092-4388/07/5002-0375
Objective: To show how tongue-to-palate contact patterns develop from childhood to
adulthood. Method: Forty-eight participants from six years to adult were involved in this
study. Participants were divided into four groups according to age with six males and six
females in each group. Prior to testing, each participant wore the EPG device until their
speech was normal as determined perceptually by the examiner. Participants then
repeated words with the following sounds in the initial position: /t/, /s/, /k/, /l/, /kl/, and
/st/. Acoustic and EPG data were collected and analyzed for each speaker. Results:
Overall representative frames of maximum contact for the phonemes /t/, /l/, and /s/
showed that the areas of most tongue-to-palate contact for each age group moved more
anteriorly as the age of the group increased. Additionally, the older groups had more
consistent contact patterns across speakers, and made contact with fewer electrodes for
each sound. Overall representative frames of maximum contact for /k/ were about the
same for all age groups. The position of the tongue on the palate for /k/ in the younger
speakers was more posterior and the tongue showed complete closure on fewer rows,
with less midline contact. When collapsed over age and sound, females showed less
contact overall, but these results were not consistent. Conclusion: As children mature into
adulthood, the tongue-to-palate contact decreases, and the place of articulation moves
forward in the mouth for consonants like /t/, /s/, and /l/, and some consonants stabilized
as the children grew older. This indicates that children continue to develop their
articulation into adulthood. Relevance to Current Study: Researchers must consider
differences between the articulation of children and adults as they apply methods
previously used to only with adults to study children.
34
Dromey, C., Nissen, S., Nohr, P., & Fletcher, S. G. (2006). Measuring tongue movements during speech: Adaptation of a magnetic jaw-tracking system. Speech Communication, 48, 463-473. doi:10.1016/j.specom.2005.05.003
Objective: The purpose of this study was to determine if a system that typically uses
magnetic pellets to track jaw movement can be adapted to effectively track the tongue
during speech. Method: The authors used an adapted version of the JT-3 jaw-tracking
device made by BioResearch Associates. Adaptations included a smaller magnetic piece
that would fit on the tongue; and the ability to control the type and frequency of recorded
data. The authors studied many functions of the device; however, most relevant is the
effect of the presence of the lingual magnet on speech. Five English speakers said words
containing /s/ and /ʃ/ in the initial, medial, and final positions in the carrier phrase, “I say
the word ___ again.” The authors compared characteristics of duration and spectral mean
for sounds produced with and without the magnet. Results: Results of the speech analysis
showed no significant difference between the fricatives produced with or without the
magnet. Conclusion: The magnet did not disrupt fricative production. Relevance to
current study: Devices involving magnetic sensors do not require the same adaptation
time as EPG devices.
Fletcher, S. G. (1989). Palatometric specification of stop, affricate and sibilant sounds. Journal
of Speech and Hearing Research, 32, 736-748. doi:10.1044/jshr.3204.736
Objective: To study how children make stop, affricate, and sibilant sounds. Method: Nine
children ages 6.8 to 14.8 were fitted with a 0.5 mm EPG sensor. Stops /t/, /d/, /k/, and /g/,
sibilants /s/, /z/, and /ʃ/, and affricates /tʃ/ and /dʒ were combined with the vowels /i/ and
/ɑ/. After a twenty-minute adjustment period, the participants were asked to said each CV
combination and to open their mouth wide between each word. Each participant was
tested five times separated by short, three- to five-minute breaks. Each sound was
separated into three segments: rising tongue-to-palate contact (segment 1), consonant
production (segment 2), and vowel production (segment 3). Results: Some of the primary
results for sibilants, showed the tongue-to-palate contact was highest in the middle of the
sibilant. Contact decreased before the following vowel and stopped decreasing as soon as
the vowel was reached. Thus, there was more tongue-to-palate contact for /i/ than for /ɑ/.
Affricates, on the other hand, had a plateau in tongue-to-palate contact in the middle of
35
the sound. The length of the plateau was equal to the length of time between the peak
contact and noise burst in sibilants. Additionally, more sensors were contacted fo
affricates than fo sibilants. A study of /ʃ/ as a phoneme and as a portion of the affricate
/tʃ/ showed that both forms of /ʃ/ had similar contact patterns. Also notable was that
affricates and stops had different vowel onset times. In relation to the other phonemes
tested, the authors found that /s/ and /z/ were formed with a groove that was narrower and
more anterior than the grooves in ʃ and the fricative portions of tʃ and dʒ. Analysis of age
differences showed that older speakers formed phonemes faster and with more precision
than the younger participants. Conclusions: From these results, the author concluded that
the plateau in the affricate is the point where the sound changes from a stop to an
affricate. Additionally, stops and affricates are indeed two different classes of consonants
with the affricate requiring more skill and effort from the speaker. Finally, as children
develop, they are able to form speech more efficiently and precisely. Relevance to
Current Study: This study is valuable in that it studies EPG use in children. The patterns
found will be important for the current researcher to consider when selecting sounds and
comparing results of children of different ages.
Fletcher, G. S., McCutcheon, M. J., & Wolf, M. B. (1975). Research note: Dynamic palatometry. Journal of Speech and Hearing Research, 18, 812-819. doi:10.1044/jshr.1804.812
Objective: A research note describing an electropalatometer. Method: The palatometer is
made of the pseudopalate and 48 electrodes. The design of the pseudopalate is two plastic
sheets with the wiring for the electrodes in between. The wires protrude through holes on
the inferior sheet so they can contact the tongue. When the tongue contacts the palate, it
creates a circuit from a 100 mv charge on the client’s wrist. This analog signal travels out
of the wires and is converted to digital information and then projected on a light emitting
diode (LED) display. The patterns from the electrodes could be recorded on magnetic
tape and then reviewed in conjunction with an acoustic spectrum, or as a summary of the
data across repeated syllables. Such a summary can help account for the variance in
repetitions of the same syllable. Conclusions: The palatometer can help measure tongue-
to-palate contact during speech, and provide researchers with a way to analyze
articulation through digital recordings and displays of palatometer measures. Relevance
36
to current study: This research note describes the design and function of the
electropalatometer, the device that will be used to gather data in the current study.
Gay, T., Lindblom, B., & Lubker, J. (1981). Production of bite-block vowels: Acoustic
equivalence by selective compensation. Journal of the Acoustical Society of America, 69, 802-810. doi: 10.1121/1.385591
Objective: To study articulator position and acoustical characteristics of vowels with and
without a bite block, and to determine if adaptation to the bite block is a
neurophysiological phenomenon. Method: Five males said a series of Swedish vowels
with and without a bite block in place. A 22.5 mm bite block was used for vowels that are
produced with a small mouth opening, and a smaller, 2.5 mm bite block for vowels that
are naturally more open. Each speaker repeated the vowel nine times and held the last
repetition. While the speaker held the vowel, the experimenters took an x-ray from the
side of the person’s face. These images were traced to show important structures for
articulation, and then digitally analyzed for cross-dimensions. The experimenters also
used a computer simulator to test how changes in articulator placement would affect the
acoustic characteristics of the vowel. Results: X-ray analysis showed that when a bite
block was in place, the articulators would compensate, or super shape to reach normal
speech placement. Articulators came closest to normal-speech positioning in areas that
were most constricted for that particular sound; more variation was seen in the more open
areas of the vocal tract. The simulations showed that areas of high constriction are the
most important in making a phoneme sound normal when a bite block is in place.
Conclusion: Vowel formation is a neurophysiological phenomenon in which vowel
formation is coded according the most important locations for an acoustically-sound
production (i.e., high constriction points). Relevance to Current Study: The results of this
study could possibly apply to an EPG device as well. That is, tongue positioning as
recorded on the EPG system may most closely resemble speech in areas of highest
constriction.
37
Gibbon, F. E., & Paterson, L. (2006). A survey of speech and language therapists’ views on electropalatography therapy outcomes in Scotland. Child Language Teaching and Therapy, 22, 275-292.
Objective: To learn about clients who participated in EPG therapy between 1993 and
2003, their therapy, and their progress as reported by their speech therapists. Methods:
The authors surveyed ten speech-language therapists in Scotland. Therapists were asked
to provide information on clients who received therapy with an EPG device. This
information included demographics, type of disorder, therapy given, and effect of EPG
therapy on progress. Results: The phoneme /s/ was treated the most in EPG therapy.
Students with functional disorders appeared to benefit least from EPG therapy, while
clients with cleft-palate seemed to benefit the most. Results also showed that 88% of the
clients had trouble generalizing skills learned in therapy. This indicates that EPG therapy
was most effective in establishing proper speech sound production, but not in
generalizing or maintaining these productions outside the clinic. Conclusions: This study
shows a need for further research in increasing generalization and maintenance in EPG
therapy. Relevance to Current Study: This article shows that EPG is effective in teaching
proper speech sound formation. This provides a need for research on how children adapt
to an EPG device to help therapists most effectively use EPG systems to help children in
therapy.
Hamlet, S. L., & Stone, M. (1982). Speech adaptation to dental prosthesis: The former lisper.
The Journal of Prosthetic Dentistry, 47, 564-478. doi:10.1016/0022-3913(82)90311-0
Objective: To determine if a history of lisping is related to difficulty adapting to a dental
retainer. Method: Participants were 13 college students who reported lisping on /s/ as
children, but now had normal speech. They were each given a dental prosthesies and
asked to wear it for two weeks. Data were before and after the two-week period to
measure jaw movement and tongue placement. In the initial measurements, speakers read
target sentences first with a 1mm electropalatography (EPG) device to measure near-
normal speech. Then participants read the sentences with an EPG the same size as the
dental prosthesis. Data for this study focused on tongue contact for initial /s/ and /z/, and
jaw height for initial /s/ and /z/ as well as /t/, /d/, /n/, and /l/. Following the two-week
38
adaptation period, participants rated how well they felt their speech had adapted to the
device. Results: Seven of the participants reported that they had adapted to the speech
device within the two-week period. The other six reported that they still felt the device
obstructed their speech, or that listeners noted a difference in their articulation. Both
participants who adapted and participants who did not adapt showed differences in place
of articulation as compared to an earlier study of normal speakers. Nonadapters made
adjustments such as tongue retraction or advancement, groove narrowing, and jaw
placement with initial placement of the prosthesis, and maintained these same adjustment
throughout the two-week period. However, participants who adapted to the device,
showed a change in their articulation adjustments between the initial and final tests.
Conclusion: People with a history of lisping use changes in tongue-palate placement (i.e.,
retracting and advancing) to adjust to the dental prosthesis more than people without a
history of lisping. Additionally, participants who were able to adapt to the dental
prosthesis, used different compensatory behaviors than those who did not. Relevance to
current study: The results of this study should be taken into account when assessing
adaptation in the present study. Researchers should be aware that adaptation time could
be longer for children with a history of lisping. Additionally, variation among speakers
may occur as a results of speakers’ different adaptation behaviors.
Jongman, A., Wayland, R., & Wong, S. (2000). Acoustic characteristics of English fricatives.
Journal of the Acoustical Society of America, 108, 1252-1263. doi: 10.1121/1.1288413
Objective: To determine which acoustic cues relate best to place of articulation; then to
describe the characteristics of these cues and their location within a sound. Method:
Twenty participants said the consonants /f/, /v/, /ɵ/, /ð/, /s/, /z/, /ʃ/, and /ʒ/ in consonant-
vowel-consonant (CVC) combinations. Their productions were recorded and the
examiners analyzed spectral properties, transition information, and noise duration for
each of the productions. All of the results from this analysis were used as predictors in a
discriminant analysis to determine which acoustic measures best predicted the place of
articulation for the fricatives. Results: Spectral peak location, normalized amplitude, and
relative amplitude were the best cues for place of articulation across all of the fricatives.
Conclusion: All four places of articulation for fricatives can be identified through
39
acoustical analysis regardless of voicing, surrounding vowels, and differences in
production. Relevance to Current Study: Characteristics of fricatives in normal speech
can serve as indicators of adaptation when speakers are tested while wearing an EPG
device.
Mantie-Kozlowski, A., & Pitt, K. (2014). Treating myofunctional disorders: A multiple-baseline
study of a new treatment using electropalatography. American Journal of Speech-Language Pathology, 23, 520-529. doi:10.1044/2014_AJSLP-14-0001
Objective: To determine if electropalatography (EPG) can be used to improve
swallowing patterns in people with non-speech orofacial myofunctional disorders
(NSOMD). Method: Three clients with NSOMD were selected for this study. This was a
multiple-baseline study in which baseline and follow-up measures consisted of recording
a series of saliva swallows on the EPG system. Each participant wore the EPG palate for
a 30-minute adaptation time prior to each testing session. Tongue-to-palate contact
patterns for each of the four stages of swallowing (pre-propulsion, propulsion, post-
propulsion, and release), were analyzed for average duration and how often the
participant’s performance matched peers who do not have NSOMD. Treatment goals
were set based on the baseline data. Treatment was held twice a week for 30-minute
sessions and included biofeedback and clinician modeling with the EPG system.
Participants were re-tested in a follow-up test five to eight weeks after treatment ended.
Results: Two of the three participants reached all of their treatment goals for the
intervention. Follow-up testing showed that two of the three participants performed above
baseline, indicating that the effect of intervention remained after five to eight weeks.
Conclusions: This study shows that EPG can be a valuable tool in helping people with
NSOMD develop better tongue-to-palate contact and lingual control in swallowing.
Relevance to Current Study: A particularly relevant part of this article is that the
experimenters allowed an adaptation time before testing the participants’ swallow.
Additionally, results from the current study may be expanded beyond articulation
therapy.
40
McAuliffe, M. J., & Cornwell, P. L. (2008). Intervention for lateral /s/ using electropalatography (EPG) biofeedback and an intensive motor learning approach: A case report. International Journal of Language and Communication Disorders 43, 219-229. doi: 0.1080/13682820701344078
Objective: To evaluate the adaptation time needed before an EPG device can be used for
therapy or data collection. Method: Participants for this study were eight female college
students. Participants were fitted with a 1-2 mm practice palate—an EPG without
electrodes. Participants used the phrase a [CVC] as a context to test the consonants /t/,
/k/, /s/, and /ʃ/ with the vowels /i/, /a/, and /u/. They said each word five times in each of
four testing sessions: (1) before using the palate, (2) immediately after inserting the
palate, (3) 45 minutes post-insertion, and (4) three hours post-insertion. Seven female
college students rated the intelligibility of the initial consonant in the third repetition of
each target word. The data were acoustically analyzed for the segment duration, first and
second formant frequencies, and consonant spectra. Results: The perceptual analysis
showed an increase in the judges’ rating of imprecision between the test immediately
before and after placement. However, ratings decreased significantly (indicating more
precision) between the tests 45 minutes and three hours following palate placement. Of
note, the phoneme /s/ adapted the most quickly as manifest by an approximately 50
percent decrease in ratings between the second and third test intervals, and a return to
near-normal precision by three hours following placement. Most phonemes to showed no
significant difference in segment duration for any of the test sessions. No differences in
vowel formant frequencies across the test sessions were found. For analysis of the first
spectral moment (mean: M1), /k/ had a significantly higher M1 three hours
postplacement as compared to the normal speech conditions. Additionally, M1 for /s/ was
lower for the three tests following placement of the practice palate. Conclusions: The
results of this study indicate that initial imprecision can be detected immediately after
insertion of the pseudopalate; however, adaptation typically occurs within 45 minutes to
three hours postplacement. Additionally, acoustic measures showed that the segment
duration of consonants and the frequency of vowel formants are not significantly affected
by an EPG palate. Relevance to current study: This study employs similar methods to
those that will be used in this thesis. These methods include testing consonant phrases at
increments following the placement of an EPG device.
41
McFarland, D. H., & Baum, S. R. (1995). Incomplete compensation to articulatory perturbation.
Journal of the Acoustical Society of America, 97, 1865-1873. doi: 10.1121/1.412060
Objective: To study the amount of compensation speakers made when speaking with a
bite block in their mouth, and to determine the effect of auditory and sensory feedback on
that compensation. Method: The participants in this study were 15 French-speaking
women. They were asked to say the vowels /i/, /a/, and /u/ in isolation and following the
consonants /p/, /t/, /k/, /s/, and /ʃ/. During each test, each stimulus sound was presented 10
times in a random order. The study included two subtests. First, the immediate subtest
which tested each participants’ normal speech and speech with a small bite block (2.5
mm for vowels and 5 mm for consonants) and a large bite block (22.5 mm for vowels and
10 mm for consonants). The second test was the post-conversation subtest in which the
speaker spoke for 15 minutes while wearing a 10 mm bite block before saying the test
stimuli. Recordings of the tests were analyzed for sound duration, vowel formant
frequency, and consonant centroid frequency. Results: There was no significant
difference for the duration of vowels or stops in the bite block condition as compared to
the normal speech condition; however, the bite block did affect the length of fricative
production. Spectral analysis showed that for vowels, F1 was highest in the large bite
block (LBB) condition. F2 frequencies differed, being higher in the LBB condition for
/u/, but lower for /i/. During consonant production, speakers’ centroid frequency was the
lowest in the LBB condition compared to the normal and small bite block (SBB)
condition. Fricatives had lower centroid frequency for both SBB and LBB tests. In post
conversation testing, the formant frequency and duration of vowels was not significantly
different from the normal-speaking condition. However, in consonant production,
significant differences were found between the post-conversation bite block condition,
and the normal-speaking condition. Conclusions: For vowels, speakers can learn to
compensate for the bite block as they use it longer. This may be a result of auditory
feedback that allows the speaker to correct for errors. Consonants, on the other hand,
require a much longer adaption time if the speaker is to adapt at all. Relevance to Current
Study: This study shows that an obstruction in the oral cavity can affect speech and that
speakers require adaptation time to compensate for the obstruction.
42
McFarland, D. H., Baum, S. R., & Chabot, C. (1996). Speech compensation to structural
modifications of the oral cavity. Journal of the Acoustical Society of America, 100, 1093-1104. doi: 10.1121/1.416286
Objective: To study the way in which speakers adapted to thick and thin pseudopalates
and to compare results to a previous study done with bite blocks. Method: Fifteen,
French-speaking women participated in this study. Each participant was fitted with a
thick (6 mm) and thin (3 mm) pseudopalate. Participants took part in two subtests. The
first was the immediate compensation subtest where speakers were tested with no palate,
and both the thin and thick palates. The second subtest was the post-conversation subtest
where speakers spoke for 15 minutes while wearing the thick pseudopalate before being
tested. Test stimuli included the vowels /i/, /a/, and /u/ in isolation, and these vowels
following the consonants /p/, /t/, /k/, /s/, and /ʃ/. Participants said each stimulus item five
times for each test condition. Recordings were analyzed for consonant and vowel
duration, first and second vowel formant frequency and consonant centroid, skewness
and kurtosis. Vowel and consonant sounds were isolated for perceptual analysis in which
judges selected the sound from several choices, and rated its quality. Results: Vowels did
not differ acoustically or perceptually between tests done with or without the palate.
Centroid, skewness, and kurtosis measures were lower with the palate than without the
palate for /s/. Identification and quality ratings were lower for the fricatives in the
immediate compensation subtest with the palate. However, in the post-conversation test,
only quality was rated lower. Stops showed some acoustic differences between palate,
and no-palate conditions during both subtests; in the immediate compensation subtests,
differences only occurred between the palate and no-palate conditions. No differences
were measured in perception ratings for consonants. Conclusion: First, vowels are the
least affected by a pseudo palate, followed by stops, and then fricatives. Second, the
subtle changes required by a thin palate may be more difficult for the mouth to
immediately adjust to than the large changes caused by the thick palate. Third, adaptation
can occur over time for some phonemes as a result of sensory feedback. Finally, people
adapt to pseudopalates differently than bite blocks. Relevance to Current Study: This
study supports the idea that people can adapt to an EPG palate over time. Additionally,
43
the current authors should keep in mind that this compensation may not occur in the same
way for all phonemes.
McLeod, S., & Searl, J. (2006). Adaptation to an electropalatograph palate: Acoustic,
impressionistic, and perceptual data. American Journal of Speech-Language Pathology, 15, 192-206. doi:10.1044/1058-0360(2006/018)
Objective: To study adaptation to an EPG in relation to acoustical and perceptual data, as
well as participant report. Method: Seven Australian adults were the subjects for this
study. Each was fitted with an EPG and pseudo-EPG device. Testing consisted of 15
trials over two days. In each test, the participant said the phrases [ə ti], and [ə si], counted
to 20, and read “The Rainbow Passage.” Acoustic measures were taken for the two
phrases. A speech-language pathologist listened to the recordings of the participants
counting and reading the passage. She was asked to record whether or not she thought the
speaker was wearing the palate, and then rate the naturalness and distortion of the
participant’s speech. Participants were periodically asked to rate the effect of the palate
on five areas: comfort, speech, tongue movement, sensation in mouth, and appearance.
Results: Both the EPG and pseudo-EPG palates affected speech. General adaptation
occurred for /t/ after about one hour, and /s/ after two hours; however, no speaker
produced speech that acoustically matched the no-palate condition. Adaptation did not
continue if the palate was removed and then reinserted. Overall, the palate had only a
small effect on perceptual distortion ratings with affricates and fricatives being most
notably affected. Participants indicated that the palate significantly affected the five areas
listed above. Some also indicated that the palate changed the sensation in their mouths
even after the palate was removed. Conclusions: Typical adults showed some ability to
adapt to an EPG device. The EPG did not grossly affect the perceptual characteristics of
speech. Relevance to Current Study: This study suggests an adaptation period of one to
two hours; results from the current thesis can be compared to this time. Additionally, the
participant’s perspective on wearing the EPG will help the current authors to anticipate
possible problems and discomforts the participants might experience during this study.
44
Nissen, S. L., & Fox, R. A. (2009). Acoustic and spectral characteristics of young children’s fricative production. Journal of the Acoustical Society of America, 126, 1369-1378. doi: 10.1121/1.3192350
Objective: This study analyzed the acoustic characteristics of the production of /p/, /t/,
and /k/ by children ages three to five, and adults. The study also discussed how the age
and gender of the speaker, place of articulation, and vowel context affected the amplitude
and spectral properties of the phonemes /p/, /t/, and /k/. Method: The experimenters
studied four groups of ten participants: (1) children age 3;0 to 3;11 years, (2) children age
4;0 to 4;11 years, (3) children age 5;0 to 5;11 years, and (4) adults. Participants were
asked to say several words in a carrier phrase, “This is a ___ again.” These words
contained one of the target consonants (/p/, /t/, or /k/) in the initial position, followed by a
vowel. The examiner elicited the target words by having the participant name pictures of
each of the items. These responses were recorded. To analyze the samples, the
experimenters first isolated the onset and offset portions of the target consonants. They
then calculated the normalized amplitude for each stop burst, and conducted a spectral
moments analysis to find the spectral mean, variance, skewness and kurtosis. A statistical
analysis was performed to detect differences resulting from the age and gender of the
speaker, place of articulation, and vowel context. Results: For the analysis of normalized
amplitude, the authors found differences in normalized amplitude for the three different
places of articulation, and the three vowel contexts. Spectral measures included spectral
slope, mean, variance, skewness and kurtosis. For spectral slope measures, the
experiemnters found different slope values for all of the target consonants in each of the
three vowel contexts; the vowel /i/ particularly elevated the slope for a preceding /k/ or
/t/. Additionally male speakers in the 5-year old and adult groups had lower spectral slope
than the females. The analysis of spectral mean found a difference among the three places
of articulation. The vowel /i/ also increased the spectral mean for /k/, but not significantly
for /p/ and /t/. Overall, the female and child participants produced a higher mean than the
adult males. Differences in spectral mean by gender were shown for /p/ and /t/ in
participants four years old and older, and for /k/ in the five-year-old and adult groups.
Generally, the mean for /t/ and /k/ was lower in males. The spectral variance for /p/ was
found to be higher than for /t/ and /k/. Additionally, the vowel /i/ lowered the spectral
45
variance for /p/ and /k/, and raised it for /t/. Measures of spectral skewness found that the
skewness for the target consonants differed from each other. When /i/ followed the
consonant /t/, the skewness of the stop was lowered. Gender differences in spectral
skewness were found starting at age 5. The spectral kurtosis measurements were different
for /p/, /t/, and /k/, and the measures increased as the place of articulation occurred farther
back in the mouth. Conclusions: This study found that the place of articulation changes
the spectral measures for different consonants. This is particularly true for spectral
variance that can help separate /p/ from /t/ and /k/ in an acoustic analysis. Secondly, this
study showed that the vowel following a consonant affected the spectral measures of that
consonant. Additionally, the effect of vowel context on the normalized amplitude of the
preceding consonant provides support for the idea that the acoustical properties of vowel
affect the perception of a preceding consonant. As for the effects of age and gender on
articulation, this study suggests that gender differences in articulation are the result of
learned articulatory patterns, not the vocal mechanism. Relevance to Current Study: The
current study uses only one vowel context with the target phonemes. According to this
article, the vowel context may change the patterns seen in speakers’ adaption to the EPG
sensor.
Searl, J., Evitts, P., & Davis, W. J. (2006). Perceptual and acoustic evidence of speaker
adaptation to a thin pseudopalate. Logopedics Phoniatrics Vocology, 31, 107-116. doi: 10.1080/14015430500390961
Objective: This study used acoustic and perceptual measures to study the effect of a thin
(0.5 mm) pseudo palate on adult production of the phonemes /t/ and /s/. Method: Eleven
adults (five male and six female) participated in the study. Each participant was fitted
with a custom pseudopalate that was 0.5 mm thick in all areas. The target phonemes /t/
and /s/ were tested by having the participants say /tik/ and /sik/ five times using the
carrier phrase a ___ again. These repetitions were spoken in a random order as prompted
by slides on a computer screen. Participants performed this task at nine different intervals
during the tests: before wearing the pseudopalate; directly after inserting the palate; at 15
minute intervals up to 60 minutes following placement of the pseudopalate; after two
hours of wearing the palate; after removing the palate; and 15 minutes after removal.
Participants did not talk between tests. Each test was recorded and acoustically analyzed.
46
Additionally, 10 experienced speech-language pathologists listened to the recordings and
performed a perceptual analysis. Participant results for the acoustic analysis and the
perceptual analysis were combined for each of the nine testing stages and a statistical
analysis was done for each. Results: The tests had three main findings. First, as manifest
by the following acoustic measures, the production of the target phonemes had acoustic
changes in the participant’s speech shortly after inserting the pseudopalate. For measures
of spectral moment 1 (SM1), the experimenters noted that SM1 for /t/ was higher 15
minutes after wearing the pseudopalate than for all other tests. For /s/, SM1 was higher
immediately after placement, and 15 and 30 minutes after placement. The stop-gap
duration was higher immediately after inserting the pseudopalate, and after 15 minutes of
wearing it. Differences in fricative duration, however, were not statistically significant.
Secondly, although these differences were observed in acoustic measures, they were not
manifest in the perceptual measures. Listeners identified the target phoneme with at least
98 percent accuracy for every test, and mean distortion ratings were between one and six
percent. The third finding was that speakers adapted to their pseudopalate in about 30
minutes after placement. This is manifest by the acoustic measures returning to
preplacement levels. Conclusion: Little or no adaptation time is needed for people to
produce perceptually undistorted consonants with a 0.5 mm pseudopalate. Acoustically
accurate productions occur after the participant has worn the pseudopalate for about 30
minutes. Relevance to Current Study: This study tests consonant phrases at increments
following the placement and removal of an EPG device, much like the methods used in
the current thesis.